|
📖 Read In Depth
|
Claude Opus 4.7
Anthropic releases Claude Opus 4.7 with 'adaptive thinking' — a new mode that dynamically allocates reasoning compute without a fixed budget. Notable API-breaking changes: updated tokenizer (1.0–1.35x token inflation), and reasoning tokens no longer included in human-readable output by default. Simon Willison flagged the adaptive thinking docs as confusing for developers migrating from prior thinking-budget APIs.
hn/Best Stories
|
Codex for almost everything
OpenAI is pushing Codex hard as a general-purpose agentic coding system — 'for almost everything.' The HN comment thread is worth skimming: critics note Claude Desktop and Cowork already do most of what's being claimed, raising the question of whether this is genuine capability leap or positioning. Also notable: a separate writeup shows Codex independently hacked a Samsung TV, suggesting real autonomous capability.
hn/Best Stories
|
Qwen3.6-35B-A3B: Agentic coding power, now open to all
Alibaba releases Qwen3.6-35B-A3B, a 35B sparse MoE model using only 3B active parameters, positioned for agentic coding. Simon Willison ran it locally in LM Studio at ~21GB and found it beat Claude Opus 4.7 on a drawing task. Open weights, runs on a laptop — this is the local model story worth paying attention to right now.
hn/Best Stories
|
The future of everything is lies, I guess: Where do we go from here?
Kyle Kingsbury (aphyr, of Jepsen fame) writes about the epistemic crisis AI-generated content is creating — not just misinformation but a deeper erosion of trust in shared facts, credentials, and expertise. This is a follow-up to a companion piece on AI's effect on jobs, and it's the kind of clear-eyed, technically grounded cultural critique that holds up. 669 HN points suggests it's landing.
hn/Best Stories
|
Open Source Isn't Dead
Cal.com is going closed source, citing AI-powered cloning as an existential threat to open-source SaaS moats. The HN thread is rich: Thunderbird's scheduling tool pledged to stay open, Simon Willison argues open source is actually MORE defensible now because AI can audit shared code at scale. A live debate about whether openness is a liability or advantage in the AI era.
hn/Best Stories
|
Cloudflare's AI Platform: an inference layer designed for agents
Cloudflare is building an inference layer explicitly designed for agents — not just model routing but stateful, multi-step task execution at the edge. Combined with their new email service for agents, this looks like Cloudflare positioning as the infrastructure layer for the agentic web. Worth understanding as a platform shift, not just a product announcement.
hn/Best Stories
|
5 Takeaways From the Live Nation Antitrust Trial
A jury found Live Nation/Ticketmaster operated as an illegal monopoly — a verdict with real structural consequences for the music industry. The five takeaways piece cuts through the legalese to explain what the jury actually found and what breakup or remedy scenarios look like. Good case study in how platform lock-in and vertical integration eventually attract antitrust fire.
nyt/Business
|
The 27-Year-Old Diplomat Waging Trump’s Cultural War With Europe
A 27-year-old political operative — five years out of college — is driving Trump's cultural confrontation with Europe, pushing allied governments on DEI, speech norms, and media funding. A sharp illustration of how ideology and youth inexperience combine in this administration's foreign policy, with real diplomatic consequences.
nyt/Top Stories
|
What Is ‘Jagged Intelligence’ and How Can It Reframe the AI Debate?
The 'jagged frontier' framing — AI is superhuman at some tasks and surprisingly bad at adjacent ones in unpredictable ways — is a useful mental model for thinking about where automation actually threatens jobs vs. where it doesn't. Worth reading for the conceptual scaffolding, not the specific job predictions.
nyt/Technology
|
|
🎬 Check It Out
|
In the Tech Heart of Texas, an Art Show Built on Data, Code and A.I.
The Blanton Museum in Austin is running a show where works change hour to hour, invite interaction, and interrogate AI's role in creativity. If you're in the Bay Area and travel to Austin, or happen to be there — this is the kind of show that sits at the intersection of code and culture in a non-gimmicky way. Check the Blanton's site for dates.
nyt/Arts
|
|
âš¡ FYI
|
Washington’s Scramble to Get Mythos, Anthropic’s Powerful New Model
Federal agencies are racing to access Claude Mythos Preview, Anthropic's new model that Anthropic says can rapidly identify — and potentially create — novel threats (biological, chemical, etc.). This signals a new phase in the government-AI lab relationship and raises serious dual-use questions around frontier model deployment.
nyt/Business
|
A Family Feud at an Oregon Winery Turns to Vinegar Over A.I. Slop
Lawyers in an Oregon inheritance dispute were fined $110,000 for submitting AI-hallucinated case citations. Courts are now hitting back hard on this practice — a useful data point on where legal liability for AI-assisted work is heading.
nyt/Top Stories
|
Carbon Removal Industry Reels as Microsoft Retreats
Microsoft has significantly pulled back from carbon removal purchasing, sending shockwaves through a nascent industry that had bet its near-term viability on big tech sustainability commitments. With data center energy demand soaring, the gap between Microsoft's climate pledges and its actual infrastructure footprint is widening publicly.
nyt/Business
|
US Bill Mandates On-Device Age Verification
A US bill would require on-device age verification, shifting the privacy/liability burden from platforms to device OS makers and app stores. If passed, this would force Apple and Google to build identity infrastructure into their platforms — a significant architectural and civil liberties inflection point.
hn/Best Stories
|
Cloudflare Email Service
Cloudflare launched an email service designed specifically for AI agents — letting agents send, receive, and act on email as a first-class capability. Paired with their AI platform announcement, this rounds out a picture of Cloudflare building the plumbing for autonomous agents operating in the real world.
hn/Best Stories
|